JP2018169769A - Sign detection system and sign detection method - Google Patents

Abstract

The present invention provides a system for performing predictive detection in consideration of aging of a plant. A predictive detection system includes a data acquisition unit that acquires device operation data and operation history data indicating the operation history of the device, operation data, operation history data, and parameters to be monitored by the device. An estimation model for estimating the value of the parameter at a time corresponding to the driving history data, an estimation unit for calculating an estimation value of the parameter based on the estimation model, an estimation value of the parameter by the estimation unit, and a data acquisition unit A state evaluation unit that evaluates the state of the device based on a deviation from a measured value or an estimated value of a parameter included in the operation data. [Selection] Figure 2

Description

  The present invention relates to a sign detection system and a sign detection method.

  In a facility such as a power plant, a model such as ANN (Artificial Nural Network) may be used to monitor the operation state of equipment. For example, in Patent Document 1, the current state of equipment is simulated by the first ANN, the state of equipment during normal operation is simulated by the second ANN, and the output data of the first ANN and the second ANN It is described that the presence or absence of a failure is determined from the difference.

US Patent Application Publication No. 2013/0318018

  By the way, in general, monitoring using a model described in Patent Document 1 or the like compares a normal model constructed by learning a process value indicating a normal state with a process value acquired in a monitored plant or the like. Anomalies are often detected when actual process values deviate from the normal model. In such a case, data used for construction of a normal model is generally past data. However, since an actual plant has a secular change, unless the normal model is updated, the normal model deviates from the normal state of the current plant, and the monitoring accuracy based on the normal model deteriorates.

  Accordingly, an object of the present invention is to provide a sign detection system and a sign detection method that can solve the above-described problems.

  According to the first aspect of the present invention, the data acquisition unit that acquires the operation data of the device and the operation history data indicating the operation history of the device, the operation data, the operation history data, and the device An estimation model for estimating a value of the parameter at a time corresponding to the driving history data for a parameter to be monitored, an estimation unit for calculating a first estimated value of the parameter, a first of the parameters A state evaluation unit that evaluates the state of the device based on a deviation between one estimated value and a measured value or a second estimated value of the parameter included in the operation data acquired by the data acquisition unit; System.

  According to the second aspect of the present invention, in the sign detection system, the operation history data may include at least one of the number of activations and the number of stops of the device.

  According to a third aspect of the present invention, in the sign detection system, the operation history data includes an operation time of the device, an operation time for each operation pattern of the device, and from the introduction of the device. At least one of the elapsed times may be included.

  Further, according to a fourth aspect of the present invention, in the predictive detection system, the operation history data includes a number of times that the maintenance / inspection operation has been performed on the device, and the maintenance / inspection operation has been performed on the device. At least one of the elapsed time from the time may be included.

  According to the fifth aspect of the present invention, in the sign detection system, the operation history data may include a cumulative value of the output of the device.

  According to a sixth aspect of the present invention, in the sign detection system, the operation data may include at least one of alarm information and event information generated by a device that monitors the device. Good.

  According to a seventh aspect of the present invention, in the sign detection system, the estimation model includes operation data of other devices of the same type as the device in addition to operation data and operation history data of the device. And it may be constructed based on driving history data.

  According to an eighth aspect of the present invention, the predictive detection system constructs the estimation model based on the operation history data and the operation data of the device at a time corresponding to the operation history data. The estimation model construction part which performs is further provided.

  Moreover, according to a ninth aspect of the present invention, in the predictive detection system, the state evaluation unit further acquires the driving history data, and the monitoring target according to the driving performance indicated by the driving history data The state of the device may be evaluated based on the parameter evaluation criteria.

  According to a tenth aspect of the present invention, in the sign detection system, the state evaluation unit is based on the deviation and the correspondence between the deviation history and the abnormality that actually occurred. Predict abnormalities that will occur in the future.

  According to the eleventh aspect of the present invention, the data acquisition unit for acquiring the operation data of the device and the operation history data indicating the operation history of the device, the operation data, and the monitoring target of the device Based on an estimation model for estimating a parameter value, an estimation unit that calculates a first estimation value of the parameter, the first estimation value of the parameter by the estimation unit, and the operation acquired by the data acquisition unit Based on the deviation from the measured value or the second estimated value of the parameter included in the data and the evaluation criterion of the parameter to be monitored according to the operation performance indicated by the operation history data, the state of the device is determined. It is a sign detection system provided with the state evaluation part to evaluate.

  According to a twelfth aspect of the present invention, the sign detection system acquires the operation data of the device and the operation history data indicating the operation history of the device, the operation data, and the operation history data. Calculating a first estimated value of the parameter based on an estimation model for estimating a value of the parameter at a time corresponding to the operation history data for the parameter to be monitored by the device; And a step of evaluating the state of the device based on a deviation between the estimated value and the measured value or the second estimated value of the parameter included in the operation data.

  According to the present invention, it is possible to detect a sign in consideration of aging of a plant or the like.

It is a figure which shows an example of the plant which monitors using the precursor detection system which concerns on this invention. It is a functional block diagram of the sign detection device in the first embodiment according to the present invention. It is a figure explaining the sign detection process in 1st embodiment which concerns on this invention. It is a flowchart which shows an example of the construction process of the estimation model in 1st embodiment which concerns on this invention. It is a flowchart which shows an example of the sign detection process in 1st embodiment which concerns on this invention. It is a figure explaining the other example of the sign detection process in 1st embodiment which concerns on this invention. It is a figure explaining the prediction of the occurrence of abnormality by the sign detection device in the first embodiment according to the present invention and the effect thereof. It is a functional block diagram of the precursor detection apparatus in 2nd embodiment which concerns on this invention. It is a figure explaining the sign detection process in 2nd embodiment which concerns on this invention. It is a flowchart which shows an example of the construction process of the determination model in 2nd embodiment which concerns on this invention. It is a figure explaining the other example of the sign detection process in 2nd embodiment which concerns on this invention.

<First embodiment>
Hereinafter, a sign detection system according to a first embodiment of the present invention will be described with reference to FIGS.
It is a figure which shows an example of the plant which monitors using the precursor detection system which concerns on this invention.
The gas turbine plant to be monitored shown in FIG. 1 includes a gas turbine 10, a generator 15, and a device 20 that controls and monitors the operation of the gas turbine 10. The gas turbine 10 and the generator 15 are connected by a rotor 14. The gas turbine 10 includes a compressor 11 that compresses air to generate compressed air, a combustor 12 that burns fuel gas in the compressed air to generate high-temperature combustion gas, a turbine 13 that is driven by the combustion gas, It has. The combustor 12 is connected to each fuel supply device (not shown) for each of the systems A, B, and C that supply fuel to the combustor 12. Between the fuel supply device and the combustor 12, fuel flow rate adjusting valves 16A, 16B, and 16C for adjusting the flow rate of the fuel in each of the systems A to C are provided. The device 20 is a control device composed of one or a plurality of computers. The device 20 adjusts the opening of an IGV (IGV: Inlet Guide Vane) 17 to adjust the flow rate of air flowing into the compressor 11 and the fuel gas to the combustor 12 by adjusting the opening of the fuel flow control valves 16A to 16C. Is controlled to drive the turbine 13 and operate the generator 15.

  Further, the apparatus 20 acquires measurement data such as temperature and pressure from sensors provided at various locations of the gas turbine 10 and the generator 15. The measurement data is taken into the gas turbine 10 and includes not only physical property data such as fuel gas and air used for actual operation but also ambient environment information such as the temperature of the operating environment. The device 20 generates a control signal for controlling the gas turbine 10 using the acquired measurement data. The measurement data includes identification information of each sensor, measurement values, measurement times, and the like. The control signal includes the time when the control signal is output. In addition, the apparatus 20 inputs measurement data to a predetermined estimation model (for example, a calculation formula for calculating a certain state quantity of the gas turbine 10), and uses the estimation value calculated by the estimation model as measurement data. Instead of or in addition to the measurement data, it may be acquired. This estimated value may be referred to as a second estimated value. These measurement data (or the second estimated value) and the control signal are collectively referred to as process data. In addition, the apparatus 20 compares the acquired measurement data with a predetermined threshold value, and if the measurement data exceeds or falls below the threshold value, an alarm indicating that the gas turbine 10 has become abnormal or has entered a state of caution. Generate information. In addition, when the gas turbine 10 is started and stopped, the operation of the equipment included in the gas turbine 10 (valve opening / closing, etc.), the change of the operation state, and the like occur, the apparatus 20 generates event information indicating the occurrence thereof. The alarm information and the event information include time information at which each information is generated. The apparatus 20 displays process data, alarm information, and event information on a display device (not shown), and an operator operates the gas turbine 10 by monitoring measurement data and alarm information displayed on the display device.

  In addition, the device 20 transmits process data, alarm information, and event information to the sign detection device 30. Regarding the process data, the apparatus 20 transmits the process data to the sign detection apparatus 30 when, for example, a predetermined change occurs in the value of the process data every time the process data is acquired. Further, the alarm information and the event information are transmitted from the device 20 to the sign detection device 30 every time the information is generated, for example. In the present embodiment, the device 20 includes a control device, a monitoring device, and the like necessary for the operation of the gas turbine 10. On the other hand, the sign detection device 30 is installed for the purpose of detecting a sign of an abnormality that will occur in the gas turbine 10 in the future and notifying the operator of this. Process data, alarm information, and event information are collectively referred to as operation data.

The sign detection device 30 detects a sign of abnormality occurring in the gas turbine 10. In particular, the sign detection device 30 according to the present embodiment performs sign detection using a prediction model that takes into account the secular change of the gas turbine 10. Next, the sign detection device 30 will be described.
FIG. 2 is a functional block diagram of the sign detection device according to the first embodiment of the present invention.
As shown in FIG. 2, the sign detection device 30 includes a data acquisition unit 31, an estimation unit 32, a state evaluation unit 33, a control unit 34, an estimation model construction unit 35, an input / output unit 36, and a storage unit 37. And comprising.
The data acquisition unit 31 acquires device operation data and operation history data indicating the device operation history. Here, the operation history data is, for example, the accumulated operation time of the gas turbine 10, the operation time for each operation pattern, the elapsed time since the gas turbine 10 was introduced and installed, the number of times of start, the number of times of stop, and the number of times of start and stop. The operation history data such as the cumulative value of the output performance of the gas turbine 10, the number of times the maintenance / inspection work has been performed on the gas turbine 10, and the maintenance history data such as the time elapsed since the maintenance / inspection work has been performed are included. The data acquisition unit 31 records the acquired operation data in the storage unit 37.
The operation history data can be calculated by using process data indicating that the gas turbine 10 is in operation, for example, the output of the generator 15, the start command signal of the gas turbine 10, and the stop command signal. For example, if it is a cumulative operation time, the apparatus 20 calculates the operation time from one start to a stop using the above process data, and accumulates this to calculate the cumulative operation time. For example, if it is the number of times of start and stop, the device 20 counts the number of start commands and stop commands output by itself, and accumulates these to calculate the number of start and stop times. The device 20 calculates the driving history data, and the data acquisition unit 31 acquires the driving history data from the device 20.
Further, the device 20 stores information on past maintenance history performed on the gas turbine 10, and the device 20 calculates the cumulative number of maintenance work, the elapsed time since the maintenance work was performed, and the like. The data acquisition unit 31 acquires maintenance history data from the device 20.

The estimation unit 32 calculates the estimated value of the parameter based on the operation data, the operation history data, and the estimation model that estimates the value of the monitoring target parameter reflecting the secular change according to the operation performance indicated by the operation history data. calculate.
The state evaluation unit 33 includes the estimated value (first estimated value) of the monitoring target parameter by the estimating unit 32, the measured value or estimated value (second estimated value) of the monitored parameter of the gas turbine 10 acquired by the data acquiring unit 31, and The state of the gas turbine 10 indicated by the monitoring target parameter is evaluated based on the deviation. For example, when the deviation is equal to or greater than a predetermined threshold, it is evaluated that the monitored parameter indicates a sign of abnormality occurring in the gas turbine 10 or a part of the equipment. Further, when detecting a sign of abnormality, the state evaluation unit 33 calculates abnormality information such as alarm information indicating that the abnormality sign has been detected, an abnormality occurrence rate, and an abnormality occurrence location.
The control unit 34 performs various controls such as activation, stop, and execution of processing of the sign detection device 30.

The estimation model construction unit 35 learns the operation history data and the operation data during normal operation of the equipment (at the time when the operation indicated by the operation history data has passed) corresponding to the operation history data, and the gas turbine An estimation model for estimating the value of the monitoring target parameter during 10 normal operations is constructed. The estimation model constructed by the estimation model construction unit 35 is a model that explicitly considers the secular change of the gas turbine 10. The estimation model is a model based on various statistical methods such as a neural network such as ANN, various machine learning, deep learning, and multiple regression analysis.
The input / output unit 36 outputs the evaluation result by the state evaluation unit 33 and the calculation result of the abnormality information to a display, another device, or the like. In addition, the input / output unit 36 receives an input operation such as instruction information to the sign detection device 30 by a monitor.
The storage unit 37 stores various data such as operation data, for example.

FIG. 3 is a diagram for explaining the sign detection process in the first embodiment according to the present invention.
The estimation unit 32 has an estimation model M. The estimation model M is based on past operation data in a normal state, and input parameters are atmospheric temperature, atmospheric pressure, relative humidity, fuel flow command value, fuel distribution ratio setting, rotation speed of the gas turbine 10, and fuel flow adjustment valve 16A. ~ 16C valve opening, fuel supply pressure for each system A ~ C, IGV opening, operation time, start / stop frequency, etc., output parameters are gas turbine output, gas turbine efficiency, cabin pressure, manifold for each fuel system For example, a neural network model is used for the pressure, the combustion vibration value of each combustor, the exhaust gas temperature from the turbine 13, and the like.
The estimation unit 32 inputs, to the estimation model M, parameters (atmospheric temperature, atmospheric pressure, fuel flow command value, etc.) necessary for estimating the monitoring target parameters (such as the output of the gas turbine 10 and the cabin pressure) among the operation data. To do. Further, the estimation unit 32 inputs operation history data such as the accumulated operation time of the gas turbine 10 and the number of start / stops to the estimation model M. The prediction model M outputs an estimated value of the gas turbine output, an estimated value of the passenger compartment pressure, and the like reflecting the secular change of the gas turbine 10 indicated by the accumulated operation time up to now and the number of start and stop times. The estimation unit 32 outputs an estimated value based on the prediction model M to the state evaluation unit 33.

  The state evaluation unit 33 acquires the estimated value of the monitoring target parameter such as the gas turbine output output by the estimation model M and the measurement value such as the current gas turbine output acquired by the data acquisition unit 31. The state evaluation unit 33 includes a subtracter S and an abnormality determination unit D. The subtracter S calculates a deviation between the estimated value obtained by the acquired estimation model M and the measured value of the corresponding parameter. For example, the subtractor S calculates a deviation between the estimated value of the gas turbine output and the measured value of the gas turbine output. Further, the subtracter S calculates a deviation between the estimated value of the passenger compartment pressure and the measured value of the passenger compartment pressure. The subtracter S outputs the calculated deviation between the estimated value of each parameter and the measured value to the abnormality determination unit D. The abnormality determination unit D compares the deviation of each parameter with a deviation threshold value determined for each parameter, and determines the presence or absence of an abnormality sign. For example, the abnormality determination unit D compares the deviation of the gas turbine output with a threshold value for gas turbine output. If the deviation of the gas turbine output exceeds a predetermined threshold value, an abnormality sign is detected for the gas turbine output. Judge that there is. Moreover, if the deviation of a gas turbine output is below a threshold value, the abnormality determination part D will determine that there is no sign of abnormality about a gas turbine output.

  The abnormality determination unit D has an abnormality information generation unit D1. The abnormality information generation unit D1 generates abnormality information such as alarm information indicating that a sign of abnormality occurrence has been detected, an abnormality location, and an abnormality occurrence rate. The alarm information includes information on the content of the abnormality and the time when it is determined that there is a sign of the abnormality. Further, the abnormality information generation unit D1 may estimate the size, accuracy, severity, and the like of the abnormality according to the magnitude of the deviation between the estimated value and the measured value, and include the information in the abnormality information. In addition, the abnormality information generation unit D1 determines in which part of the gas turbine 10 an abnormality occurs according to the type of the monitoring target parameter that exceeds the threshold value, and estimates a place where the occurrence of the abnormality is expected. Further, the abnormality information generation unit D1 calculates an abnormality occurrence rate and a failure rate of a part where a sign of abnormality is detected based on the past abnormality occurrence results, failure results, and the like. The storage unit 37 stores past failure record information. Further, the abnormality information generation unit D1 may generate abnormality information based on the values of a plurality of monitoring target parameters. For example, when both the monitoring target parameters of the gas turbine output and the passenger compartment pressure exceed the threshold, alarm information indicating that “abnormal X” is a sign may be generated.

  Predictive detection based on conventional prediction models often does not take into account plant aging, so the value of the monitored parameter estimated by the prediction model deviates from the actual state of the plant after a certain period of operation. Often it has been. On the other hand, in the present embodiment, the current output parameter value of the gas turbine 10 corresponding to the secular change is estimated based on a predetermined estimation model. Then, it is determined whether the estimated value is positive and the measured value of the output parameter of the gas turbine 10 is in a normal range. Since the abnormality is determined based on the threshold value corresponding to the current plant state, it is possible to detect the sign with high accuracy.

Next, an estimation model construction process according to this embodiment will be described.
FIG. 4 is a flowchart showing an example of the estimation model construction process in the first embodiment according to the present invention.
First, the data acquisition unit 31 acquires operation history data corresponding to the operation data together with the operation data of the gas turbine 10 (step S11). The operation history data corresponding to the operation data is, for example, the operation data such as the atmospheric temperature X, the fuel flow rate command value Y, the gas turbine 10 output Z, etc. measured by the sensor at YYYY year MM month DD day hh hour mm minute ss second Or when it is what was output by the apparatus 20, it is the cumulative operation time to the YYYY year MM month DD day hh hour mm minute ss second (or a predetermined period before and after that) and the number of start / stop times. Further, the operation data to be acquired is operation data at the time of past normal operation collected in the gas turbine 10 to be monitored. As an example of the operation history data, in addition to the cumulative operation time and the number of start / stop operations, the operation time for each operation pattern, the number of start times or the operation frequency, the elapsed time since the gas turbine 10 was installed, the gas after introduction The accumulated value (MWh) of the output by the turbine 10, the number of times the maintenance / inspection work has been performed, the time elapsed since the maintenance / inspection work has been performed, and the like may be included. The operation time for each operation pattern is, for example, the cumulative operation time in the rated operation, the cumulative operation time in the partial load operation, or the cumulative operation time when the load fluctuates. For example, the operation time for each operation pattern includes hot start (for example, stop time within 8 hours), warm start (for example, stop time within 24 hours), cold start (for example, stop time over 24 hours) and operation. When the patterns are classified, it is the cumulative operation time when the operation is started in the start mode of each pattern. In addition, for example, the number of activations for each operation pattern includes the cumulative number of activations that have been hot-started, the cumulative number of activations that have been warm-started, and the cumulative number of activations that have been cold-started. For example, the operation frequency for each operation pattern is, for example, how many times a hot start, a warm start, and a cold start are performed within a predetermined period based on the operation data collection date. Further, the elapsed time after installing the gas turbine 10 is an elapsed time including all of the operation time and the stop time after the installation of the gas turbine 10. The number of maintenance inspections is the total number of periodic inspections, part repairs and replacements, part upgrades, etc. that have been performed up to that point with reference to the date of collection of the corresponding operation data. As for the number of times of performing the maintenance inspection work, the number of maintenance inspections for each inspection place or parts for which maintenance has been performed may be used. In addition, the time elapsed since the maintenance inspection work is, for example, the time that has elapsed since the first or last periodic inspection, or the last time a part has been inspected or replaced several times. This is the time that has elapsed since the exchange.

Next, the data acquisition unit 31 records the acquired operation data and operation history data in the correspondence relationship in association with the storage unit 37 (step S12).
Next, the estimation model construction unit 35 determines whether or not operation data for a predetermined period necessary for construction of the estimation model has been accumulated in the storage unit 37 (step S13). If not accumulated (step S13; No), the processing from step S11 is repeated. When operation data for a predetermined period is accumulated (step S13; Yes), the estimated model construction unit 35 constructs an estimated model M (step S14). For example, for the operation history data β corresponding to certain operation data α, the estimated model construction unit 35 includes input parameters (atmospheric temperature, atmospheric pressure, fuel flow command value, etc.) and operation history data (cumulative operation) among the operation data α. When the time and the number of times of starting and stopping are input to the model, an estimation model that outputs monitoring target parameters (gas turbine output, cabin pressure, etc.) in the operation data is constructed by a technique such as ANN. The estimated model construction unit 35 records the constructed estimated model M in the storage unit 37. Next, a flow of online sign detection processing using the constructed estimation model M will be described.

FIG. 5 is a flowchart showing an example of the sign detection process in the first embodiment according to the present invention.
First, the data acquisition unit 31 acquires operation data of the operating gas turbine 10 (step S21). For example, the data acquisition unit 31 acquires the latest measurement values (atmospheric temperature, atmospheric pressure, output of the gas turbine 10, cabin pressure, etc.) measured by various sensors via the device 20. Further, the data acquisition unit 31 acquires the latest control signal value (IGV opening, fuel flow command value, valve opening of each fuel system, etc.) instructed by the device 20 from the device 20. The operation data acquired by the data acquisition unit 31 includes input parameters and monitoring target parameters (output parameters). Further, the data acquisition unit 31 acquires operation history data (cumulative operation time, number of start / stops, elapsed time after periodic inspection, etc.) corresponding to the acquired operation data from the device 20.
The data acquisition unit 31 outputs input parameters and operation history data among the acquired operation data to the estimation unit 32. Further, the data acquisition unit 31 outputs the monitoring target parameter in the acquired operation data to the state evaluation unit 33.

  Next, the estimation unit 32 reads the estimation model M from the storage unit 37 and inputs the operation data (input parameters) and the operation history data acquired from the data acquisition unit 31 to the read estimation model M. The estimation model M calculates an estimated value of the monitoring target parameter using the input value (step S22). The estimation unit 32 outputs the estimated value of the monitoring target parameter to the state evaluation unit 33.

  Next, the state evaluation unit 33 calculates the deviation between the estimated value of the monitoring target parameter acquired from the estimation unit 32 and the measured value of the monitoring target parameter acquired from the data acquisition unit 31 (step S23). For example, the subtractor S included in the state evaluation unit 33 subtracts the measured value of the gas turbine output from the estimated value of the gas turbine output. The subtracter S outputs the calculated deviation to the abnormality determination unit D included in the state evaluation unit 33. The subtractor S similarly calculates a deviation for the other monitoring target parameters, associates the monitoring target parameter identification information (for example, the name of the monitoring target parameter) with the deviation, and outputs it to the abnormality determination unit D.

  Next, the state evaluation unit 33 evaluates the monitoring target parameter based on the deviation between the estimated value and the measured value. For example, the abnormality determination unit D compares a predetermined threshold value determined for each monitoring target parameter with the deviation of the monitoring target parameter calculated by the subtracter S. The abnormality determination unit D determines that there is a sign of abnormality when the deviation exceeds the threshold, and determines that there is no sign of abnormality when the deviation is within the threshold.

  When it is determined that there is a sign of abnormality (step S26; Yes), the state evaluation unit 33 notifies that an abnormality sign has been detected via the output unit 104 (step S27). For example, the abnormality information generation unit D1 generates alarm information including the name of the monitoring target parameter determined to have an abnormality sign, the abnormality occurrence time, and the like. Further, for example, the abnormality information generation unit D1 specifies an abnormality occurrence location based on the monitoring target parameter, and generates information indicating the abnormality occurrence location. Further, for example, the abnormality information generation unit D1 calculates the number of occurrences of the alarm based on the past alarm information, or the abnormality occurrence rate at the abnormality occurrence point based on the past abnormality information, and indicates the abnormality occurrence rate. Is generated. Further, the state evaluation unit 33 may determine the presence / absence of an abnormality sign by using a plurality of monitoring target parameters. For example, when the deviation for the monitoring target parameter K1 exceeds the threshold value, the state evaluation unit 33 determines that it is a sign of the abnormality K1 ′, and both the deviation for the monitoring target parameter K1 and the deviation for the monitoring target parameter K2 are both. When the threshold value is exceeded, the state evaluation unit 33 may determine that it is a sign of an abnormality K2 ′. Information that associates the monitoring target parameter with the abnormality occurrence location, past alarm information, past abnormality information for each abnormality occurrence location, and the like are recorded in the storage unit 37. The state evaluation unit 33 outputs the alarm information generated by the abnormality information generation unit D1, the abnormality occurrence rate, information on the abnormality occurrence point, and the like to the input / output unit 36. The input / output unit 36 displays alarm information and the like on a display connected to the sign detection device 30.

  When it is determined that there is no sign of abnormality (step S26; No), the sign detection device 30 determines the end of the sign detection process (step S28). For example, when the monitoring person inputs a process stop command to the sign detection device 30 via the input / output unit 36, the control unit 34 determines to end the sign detection process. When it determines with complete | finishing a sign detection process (step S28; Yes), the control part 34 stops a sign detection process. In this case, this flowchart is terminated. When continuing the sign detection process (step S28; No), the process from step S21 is repeated.

  The abnormality in which the sign is detected in step S24 is, for example, an event that may be determined by the device 20 as an abnormality after several hours. Or it is an event that requires maintenance work after stopping operation in several months. According to the sign detection device 30 of the present embodiment, as the operation history data, for example, an estimation model is constructed by using the accumulated operation time, or an estimated value is calculated, so that the monitoring target parameter considering the secular change of the plant is taken into account. Can be estimated. In addition, it is conceivable that not only the cumulative operating time but also the operating load and operating conditions during that time will affect the deterioration of the equipment. For example, as the operation history data, the cumulative value of the gas turbine output, the operation for each operation pattern, etc. By building an estimation model and calculating an estimated value using time or the like, it is possible to estimate the value of the parameter to be monitored that reflects the influence of the degree of the operational load of the plant. In addition, as the operation history data, for example, the estimation model is constructed and the estimated value is calculated by using the number of start / stop times, the operation frequency, the number of maintenance operations, the elapsed time from the maintenance operation, etc. It is possible to estimate the value of the monitoring target parameter in consideration of the degree. That is, since the value of the monitoring target parameter is estimated by the estimation model M using various operation history data as parameters, it is possible to perform an estimation that more reflects the current state of the plant. Thereby, a highly accurate precursor detection becomes possible. In the above description, the case where the estimation model M is used for predictive detection has been described as an example. However, the estimation model M may be used for abnormality determination with respect to current plant operation data. Further, the state evaluation unit 33 determines not only the presence / absence of an abnormality sign but also the probability of occurrence of an abnormality, and even if there is no abnormality sign, the operation state of the plant is “good”, “normal”, You may make it notify to a monitoring person by evaluating according to steps, such as "slightly high load."

In addition to process data (measurement data and control signal), alarm information and event information generated by the device 20 may be added to the operation data as input parameters to the estimation model M.
FIG. 6 is a diagram for explaining another example of the sign detection process in the first embodiment according to the present invention.
In the modification of the first embodiment shown in FIG. 6, alarm information and event information are input to the estimation model M ′ in addition to the input parameters described in FIG. The estimation model construction unit 35 acquires input parameters including alarm information and event information, and constructs an estimation model. At this time, when using a learning method (RNN: Recurrent Neural Network or the like) for learning past time-series changes, time-series process data, alarm information, and event information can be input as operation data. In the case of a model that does not learn time-series changes, for example, the estimation model M ′ may be constructed by inputting alarm information or event information that occurred within a predetermined time.

In the case of the configuration of FIG. 6, in the sign detection process, the data acquisition unit 31 acquires alarm information and event information in addition to the input parameters described in FIG. 3, and the estimation unit 32 adds to the accumulated operation time and the like. The value of the monitoring target parameter reflecting the occurrence status of alarm information and event information is estimated. For example, even if the values of the input parameters such as the atmospheric temperature, the fuel flow rate command value, and the cumulative operation time are the same, there is a possibility that the monitoring target parameter is greatly different between the situation where the alarm information is generated and the situation where it is not. In addition, even if the input parameter values such as the atmospheric temperature, fuel flow command value, and cumulative operation time are the same, the operation (event information) of the seemingly irrelevant device has an indirect effect, and as a result, the monitored parameter May also affect the value of. With the estimation model M ′ illustrated in FIG. 6, it is possible to estimate the value of the monitoring target parameter including the relationship with alarm information and event information that have not been conscious of before.
In the above description, both alarm information and event information are used, but only one of them may be used.

  Further, it is preferable that the learning data used for the construction of the estimation models M and M ′ is large. Therefore, in addition to the operation data and operation history data collected at the monitored plant, the operation data and operation history data collected at the same kind of plant and equipment are added to the operation data of the monitored plant, and the estimated models M and M ′ Construction may be performed.

  Further, in the sign detection device 30 illustrated in FIG. 2, the case where the sign detection device 30 includes the estimation model construction unit 35 is illustrated, but a configuration without the estimation model construction unit 35 may be employed. In this case, for example, the estimated model M is constructed by another computer, and the estimated model M is recorded in the storage unit 37.

  Moreover, the sign detection device 30 can also be used for future prediction of the plant state. For example, if a transition of a value indicated by a certain monitoring target parameter can be predicted, it can be used for estimation of the deterioration progress speed and replacement time of parts related to the monitoring target parameter, planning of a maintenance plan, and the like.

FIG. 7 is a diagram for explaining the prediction of the occurrence of an abnormality and its effect by the sign detection device in the first embodiment according to the present invention.
The vertical axis of the graph shown in FIG. 7 indicates the value of the monitoring target parameter Q related to the component P used in the gas turbine 10, and the horizontal axis indicates the accumulated operation time from the start of operation of the gas turbine 10. For example, it is assumed that the gas turbine 10 has an operation plan that keeps operating at a rated load. In this case, the value of the operation data (atmospheric temperature or the like) as an input parameter to the estimation model M in the future operation can be calculated from the past operation results. Further, it is assumed that the operation history data input to the estimation model M is the accumulated operation time. Further, the current cumulative operation time is T0, and the estimated value of the monitoring target parameter Q based on the estimation model M at this time is R0. Further, it is defined that the component P must be replaced when the value of the monitoring target parameter Q is equal to or less than the threshold value R. In addition, the sign detection device 30 is configured to operate in an operation mode in which the trend prediction of the monitoring target parameter is performed in addition to the operation mode in which the sign detection is performed. In the operation mode in which the trend prediction is performed, the sign detection device 30 outputs an estimated value of the monitoring target parameter based on the estimation model M in a specified period (for example, specified by the cumulative operation time).

Under these conditions, the maintenance planner can plan the replacement timing of the parts P as follows. For example, the planner inputs to the sign detection device 30 so as to operate in the operation mode in which the trend prediction of the monitoring target parameter Q is performed. Subsequently, the planner inputs the values of the input parameters (such as the atmospheric temperature) relating to the future operation based on the past operation results and the predictive detection device 30 as a period for designating the accumulated operation time T0 to T3.
Then, the control part 34 starts the trend prediction process of the monitoring object parameter. First, the input / output unit 36 receives input of those values, outputs the values of the input parameters to the data acquisition unit 31, and outputs the accumulated operation times T0 to T3 to the control unit 34. Then, the control unit 34 adds a predetermined period ΔT to the cumulative operation time T0, and outputs the cumulative operation time T0 + ΔT after the addition to the data acquisition unit 31. The data acquisition unit 31 outputs the input parameters and the accumulated operation time T0 + ΔT to the estimation unit 32. The estimation unit 32 calculates an estimated value of the monitoring target parameter Q in the cumulative operation time T0 + ΔT. When the calculation of the estimated value is completed, the control unit 34 outputs the cumulative operation time T0 + 2 · ΔT to the data acquisition unit 31. The estimation unit 32 calculates an estimated value of the monitoring target parameter Q in the cumulative operation time T0 + 2 · ΔT. Hereinafter, similarly, the estimation unit 32 repeats the calculation of the estimated value of the monitoring target parameter Q until the cumulative operation time reaches T3 while increasing the cumulative operation time by ΔT. A graph q1 is obtained by such processing. The planner grasps from the graph q1 that when the cumulative operation time reaches T2, the monitoring target parameter Q reaches the threshold value R and the part P must be replaced. For example, if it is planned to perform periodic inspection when the cumulative operation time becomes T1, the planner can make a plan to replace the parts P at this timing. Similarly, the sign detection device 30 can also predict the value of the monitoring target parameter Q after replacement of the component P (graph q2).

<Second embodiment>
Hereinafter, a sign detection system according to a second embodiment of the present invention will be described with reference to FIGS.
The sign detection device 30A according to the second embodiment will be described. The sign detection device 30 </ b> A detects an abnormality sign by a method different from that of the first embodiment. In the first embodiment, the estimation model M estimates the monitoring target parameter in consideration of the secular change of the plant. In the second embodiment, the state evaluation unit 33A evaluates the monitoring target parameter by an evaluation method that takes into account the secular change of the plant.

FIG. 8 is a functional block diagram of the sign detection device according to the second embodiment of the present invention.
Among the configurations according to the second embodiment of the present invention, the same components as those constituting the sign detection device 30 according to the first embodiment are denoted by the same reference numerals, and description thereof is omitted. As shown in the drawing, the sign detection device 30A includes a data acquisition unit 31, an estimation unit 32, a state evaluation unit 33A, a control unit 34, an estimation model construction unit 35, an input / output unit 36, a storage unit 37, The determination model construction part 38 and the abnormality performance data acquisition part 39 are provided.
The state evaluation unit 33A acquires the operation history data in addition to the estimated value and the measurement value of the monitoring target parameter, and evaluates the state of the device based on the evaluation standard corresponding to the operation performance indicated by the operation history data.
The determination model construction unit 38 learns the estimated and measured values of the monitoring target parameters including those at the time of occurrence of the abnormality and the operation history data at that time, and evaluates according to the operation results indicated by the operation history data Build a decision model for calculating criteria.
The abnormality record data acquisition unit 39 includes the details (type and scale) of the abnormality, the location of the abnormality, the method for dealing with the abnormality, the estimated value and the measured value of the monitoring target parameter at the time of the abnormality, Information such as operation history data at that time is acquired.

FIG. 9 is a diagram illustrating the sign detection process in the second embodiment according to the present invention.
The estimation model M ′ and the subtracter S are the same as those described with reference to FIGS. That is, the estimation model M ′ is an estimation model that estimates the value of the monitoring target parameter according to the secular change based on past normal operation data. Further, the subtracter S calculates a deviation between the estimated value and the measured value of each monitoring target parameter, and outputs the deviation to the abnormality determination unit D.

  The abnormality determination unit D included in the state evaluation unit 33A of the present embodiment has a determination model N. The determination model N calculates an evaluation criterion at the time when the driving indicated by the driving history data is performed. For example, it is assumed that it is appropriate to determine that a certain parameter to be monitored is an abnormal sign when the deviation between the estimated value and the measured value is “10” at the time when the operation years are short. On the other hand, when the operating years become long, for example, when the deviation between the estimated value and the measured value becomes “10”, the deviation rapidly increases after that, unlike when the operating years are short, which is considerably faster than when the operating years are short. Suppose that an abnormality occurs at the time. In such a case, the threshold value can be changed in accordance with the accumulated operation time (operation history data) irrespective of calculating the estimated value in consideration of the secular change of the monitoring target parameter using the estimation model M ′. It is considered appropriate. In such a case, the determination model N determines that there is an abnormal sign based on a threshold value (for example, “5”) corresponding to the accumulated operation time.

  In addition, as another example of the determination model N, a model is considered in which operation data for a predetermined period is input instead of operation history data and an abnormality sign is determined. For example, it is assumed that the abnormality to be generated differs depending on the deviation pattern (deviation history) of these three monitoring target parameters for the monitoring target parameters a1, a2, and a3. The determination model N includes a history of deviation between an estimated value and a measured value of the estimated model M ′ for one or a plurality of monitoring target parameters (for example, each of the monitoring target parameters a1, a2, and a3), and a deviation of each deviation. Correspondences with abnormalities that actually occurred in the history are stored. In such a case, with respect to the parameters a1, a2, and a3, information on the deviation between the estimated value based on the estimated model M ′ and the actual measured value is recorded in the storage unit 37 for a predetermined period. In the determination model N, the subtracter S calculates each of one or a plurality of monitoring target parameters (for example, monitoring target parameters a1, a2, a3) for a predetermined period in the past with the current time recorded in the storage unit 37 as a reference. Enter the deviation time series information (deviation history). The determination model N determines an abnormal sign based on the correspondence relationship according to the input history of deviation in a predetermined period. As a result, it is possible to detect an anomaly sign in consideration of changes in the monitoring target parameter over time.

  In the above example, the driving history data is included in the input parameter of the determination model N. However, at least one of alarm information and event information may be included in the input parameter to determine the abnormality sign. Good.

Next, an example of a determination model construction process according to the present embodiment will be described.
FIG. 10 is a flowchart showing an example of a determination model construction process according to the second embodiment of the present invention.
As a premise, it is assumed that, based on which monitoring target parameter is used to determine an abnormality sign for a certain abnormality, and which time point before the abnormality occurs is determined as an abnormality sign detection time point.
First, the data acquisition unit 31 acquires operation history data corresponding to deviation data between an estimated value and a measured value based on the estimated model M (or M ′) in a predetermined period from when the abnormality occurs to when the abnormality occurs (step S31). ). Next, the data acquisition unit 31 records the acquired deviation data and operation history data in the correspondence relationship in association with each other in the storage unit 37 (step S32). Further, the abnormality record data acquisition unit 39 acquires abnormality information (abnormality type, scale, occurrence location, coping method, etc.) corresponding to the deviation data at the time of abnormality acquired in step S31 and records it in the storage unit 37 ( Step S33). Next, the determination model construction unit 38 constructs a determination model N (step S34). For example, the determination model construction unit 38 calculates the relationship between the accumulated operation time and the magnitude of the deviation when an abnormality occurs, from the relationship between the history of deviation and the accumulated operation time for an abnormality that has actually occurred. Further, the determination model construction unit 38 calculates a deviation at a time point before the predetermined period when the abnormality sign detection time is determined from the occurrence of the abnormality, and uses the value of the deviation as a threshold value for abnormality sign detection. In addition, the determination model construction unit 38 calculates the cumulative operation time at the time when the threshold is set. The determination model construction unit 38 records the calculated accumulated operation time, threshold value, and abnormality information in the storage unit 37 as a set. Thereby, the threshold value according to the accumulated operation time of a plant can be set for every kind of abnormality.

FIG. 11 is a diagram for explaining another example of the sign detection process in the second embodiment according to the present invention.
In the example shown in FIG. 11, no driving history data is input to the estimated model M ″. On the other hand, for the determination model N, operation history data and alarm information are input. As shown in FIG. 11, in general abnormality sign detection using an estimation model that does not consider secular change, driving history data is input only for the abnormality sign determination process, and an evaluation standard corresponding to secular change is used. Based on this, an abnormality sign is determined. As a result, an abnormal symptom considering the secular change of the plant becomes possible.
Note that the type of operation history data used for the construction of the determination model N and the abnormality sign determination is not limited to the cumulative operation time, and the one exemplified in the estimation model M (such as FIG. 3) of the first embodiment can be used.

  Further, in the sign detection device 30A illustrated in FIG. 2, the case where the sign detection device 30A includes the determination model construction unit 38 and the abnormality record data acquisition unit 39 is illustrated, but the function detection unit 30A may not include these function units. Is possible. In this case, for example, the determination model N is constructed by another computer, and the determination model N is recorded in the storage unit 37.

  According to the present embodiment, an abnormality sign can be performed based on an evaluation standard according to a secular change, an operational load, an operation condition, and the like of the plant. It can be changed together.

  In the first embodiment and the second embodiment described above, the sign detection is performed based on the deviation between the first estimated value and the measured value of the monitoring target parameter. Instead, the second estimated value may be used, and the sign detection may be performed based on the deviation between the first estimated value and the second estimated value of the monitoring target parameter.

  The sign detection devices 30 and 30A are examples of the sign detection system. Each process in the sign detection devices 30 and 30A is stored in a computer-readable recording medium in the form of a program, and the computer of the sign detection devices 30 and 30A reads and executes the program. The above processing is performed. Here, the computer-readable recording medium means a magnetic disk, a magneto-optical disk, a CD-ROM, a DVD-ROM, a semiconductor memory, or the like. Alternatively, the computer program may be distributed to the computer via a communication line, and the computer that has received the distribution may execute the program.

The program may be for realizing a part of the functions described above. Furthermore, what can implement | achieve the function mentioned above in combination with the program already recorded on the computer system, what is called a difference file (difference program) may be sufficient.
In addition, the sign detection devices 30 and 30A may be configured by a single computer, or may be configured by a plurality of computers connected to be communicable.

  In addition, it is possible to appropriately replace the components in the above-described embodiments with known components without departing from the spirit of the present invention. The technical scope of the present invention is not limited to the above-described embodiment, and various modifications can be made without departing from the spirit of the present invention.

DESCRIPTION OF SYMBOLS 10 ... Gas turbine, 15 ... Generator, 20 ... Apparatus, 11 ... Compressor,
12 ... Combustor, 13 ... Turbine, 14 ... Rotor, 16A, 16B, 16C ... Fuel flow rate adjusting valve, 17 ... IGV, 30, 30A ... Predictive detection device,
31 ... Data acquisition unit, 32 ... Estimation unit, 33 ... State evaluation unit, 34 ... Control unit,
35 ... Estimated model construction unit, 36 ... Input / output unit, 37 ... Storage unit, 38 ... Determination model construction unit, 39 ... Abnormal performance data acquisition unit, A, B, C ...・ Fuel system, D ... Abnormality judgment unit, M, M '... Estimated model, P ... Parts, Q ... Monitored parameter

Claims (12)

A data acquisition unit for acquiring the operation data of the device and the operation history data indicating the operation history of the device;
Based on the operation data, the operation history data, and an estimation model that estimates a value of the parameter at a time corresponding to the operation history data for a parameter to be monitored by the device, the first of the parameters An estimation unit for calculating an estimated value;
A state evaluating unit that evaluates the state of the device based on a deviation between the first estimated value of the parameter and the measured value or second estimated value of the parameter included in the operation data acquired by the data acquiring unit;
Predictive detection system with The operation history data includes at least one of the number of times the device is activated and the number of times it is stopped.
The sign detection system according to claim 1. The operation history data includes at least one of an operation time of the device, an operation time for each operation pattern of the device, and an elapsed time since the introduction of the device.
The sign detection system according to claim 1 or 2. The operation history data includes at least one of the number of times that the maintenance inspection work has been performed on the device and the elapsed time since the maintenance inspection work has been performed on the device.
The sign detection system according to any one of claims 1 to 3. The operation history data includes a cumulative value of the output of the device,
The sign detection system according to any one of claims 1 to 4. The operation data includes at least one of alarm information and event information generated by a device that monitors the device.
The sign detection system according to any one of claims 1 to 5. The estimation model is constructed based on operation data and operation history data of other devices of the same type as the device, in addition to operation data and operation history data of the device,
The sign detection system according to any one of claims 1 to 6. An estimation model construction unit that constructs the estimation model based on the operation history data and the operation data of the device at a time corresponding to the operation history data;
The sign detection system according to any one of claims 1 to 7, further comprising: The state evaluation unit further acquires the operation history data, and evaluates the state of the device based on an evaluation criterion of the parameter to be monitored according to the operation performance indicated by the operation history data.
The sign detection system according to any one of claims 1 to 8. The state evaluation unit predicts an abnormality that will occur in the future based on the deviation and a correspondence relationship between the deviation history and the abnormality that has actually occurred.
The sign detection system according to any one of claims 1 to 9. A data acquisition unit for acquiring the operation data of the device and the operation history data indicating the operation history of the device;
An estimation unit that calculates a first estimated value of the parameter based on the operation data and an estimation model that estimates a value of the parameter to be monitored by the device;
Depending on the deviation between the first estimated value of the parameter by the estimating unit and the measured value or the second estimated value of the parameter included in the driving data acquired by the data acquiring unit, and the driving performance indicated by the driving history data In addition, based on the evaluation criteria for the parameter to be monitored, a state evaluation unit that evaluates the state of the device,
Predictive detection system with The sign detection system
Obtaining operation data of the device and operation history data indicating an operation history of the device;
Based on the operation data, the operation history data, and an estimation model that estimates a value of the parameter at a time corresponding to the operation history data for a parameter to be monitored by the device, the first of the parameters Calculating an estimate;
Evaluating the state of the device based on a deviation between the first estimated value and the measured value or second estimated value of the parameter included in the operation data;
A sign detection method comprising:

Images